What Does Kubernetes Cost for Small Businesses?
In short: a small production workload with 2 vCPU, 4 GiB RAM, and 20 GiB of storage costs between about €41 and €300 per month (net), depending on the model. The most expensive part is rarely the infrastructure, it’s the working time spent running the cluster. This post works through three paths: running it yourself, AWS EKS, and a managed namespace.
Container orchestration with Kubernetes is the industry standard for deploying modern applications. But the pricing models of the major cloud providers are built for enterprise budgets, not for small and medium-sized businesses. If you’re a five-person team running three microservices on AWS EKS, you can easily pay $150 to $400 per month before the first hour of operations work.
Three Common Paths to Kubernetes
1. Self-Hosted on Your Own Servers
You rent bare-metal servers or VMs and install Kubernetes yourself, using kubeadm, k3s, or Rancher. It sounds cheap because you only pay for the servers.
What’s often overlooked:
- Time spent on setup and maintenance: Updates, etcd backups, certificate rotation, monitoring stack, all on you.
- Downtime risk: Without a redundant control plane, a cluster outage is only a matter of time.
- Hidden labor costs: Assume an experienced DevOps engineer costs €75,000 to €95,000 gross salary per year. With employer contributions, it’s closer to €90,000 to €115,000. Even dedicating a quarter of their time to cluster maintenance adds €1,900 to €2,400/month.
Assumed total cost: €150 to €300/month (net) infrastructure + significant time investment.
2. Managed Kubernetes from a Hyperscaler, Using AWS EKS as the Example
The major cloud providers handle the control plane for you. In return, you pay a base fee, plus compute, storage, and networking.
AWS EKS example for a typical SMB workload (2 vCPU, 4 GiB RAM, 20 GiB storage), region eu-central-1:
| Item | Monthly Cost (net) |
|---|---|
| EKS Control Plane | $73.00 |
| EC2 t3.medium (2 vCPU, 4 GiB) | $35.04 |
| EBS Storage 20 GiB (gp3) | ~$1.90 |
| Load Balancer (ALB) | ~$19.71 |
| Egress 50 GiB | $0.00 (100 GB/month free) |
| Total (net) | ~$129.65 (~€118) |
The $73 control plane price applies to standard support ($0.10/hour). If a cluster slips into extended support, the hourly rate jumps to $0.60, roughly $438/month for the control plane alone.
That’s the minimum configuration. High availability doubles your compute costs. Add ECR for container images, CloudWatch for logs, and the learning curve for IAM roles, VPC configuration, and dozens of AWS-specific abstractions.
Realistic for SMBs (assumed): $150 to $400/month net, plus onboarding time into the AWS ecosystem. That range is on the conservative side: a real setup with private subnets also carries a NAT Gateway (about $38/month plus $0.052 per GB processed) and charges for public IPv4 addresses.
3. Managed Kubernetes Namespace with ITSH
With ITSH, you don’t rent an entire cluster. Instead, you get an isolated namespace on an existing cluster in German data centers.
Before you pay anything, you can try the platform on the free tier: one free namespace per account, once (200 mCPU, 512 MiB RAM, 1 GiB NFS storage, 10 GiB outbound traffic per month). When it gets too small, you switch it to pay-as-you-go in the portal. That needs a payment method on file and a confirmation, and your workloads keep running without interruption. The free tier has limits, for example no custom domains and outbound traffic over HTTP and HTTPS only. The full list is in the documentation.
ITSH cost breakdown for the same workload (2 vCPU, 4 GiB RAM, 20 GiB storage, 50 GiB egress):
| Item | Monthly Cost (net) |
|---|---|
| Base fee (PAYG) | €2.52 |
| 2 vCPU (730 h) | €21.90 |
| 4 GiB RAM (730 h) | €14.60 |
| 20 GiB Storage | €1.60 |
| 50 GiB Egress | €0.50 |
| Total (net) | €41.12 |
| All prices net, plus 19% VAT | = €48.93 gross |
There are fixed tiers as an alternative: Namespace S with 1 vCPU, 2 GiB RAM, and 100 GiB of included egress for €15.00 gross per month, more than enough for many small projects. Above it sit Namespace M (2 vCPU, 4 GiB RAM, 250 GiB egress) at €29.00 gross and Namespace L (4 vCPU, 8 GiB RAM, 500 GiB egress) at €49.00 gross. No fixed tier includes storage, PVCs are billed separately.
If your load fills a fixed tier most of the time, the tier is cheaper: 2 vCPU and 4 GiB cost €29.00 gross as a Namespace M, and €46.43 gross on PAYG (base fee and compute; storage comes on top in both cases). With small or varying load, or when services sleep part of the time, PAYG is cheaper.
Cost Comparison: Self-Hosted, AWS EKS, and Managed Namespace
| Self-Hosted | AWS EKS | ITSH Namespace | |
|---|---|---|---|
| What you get | Your own cluster | Your own cluster | Namespace on a shared cluster, no custom CRDs |
| Base fee | €0 | ~€66 ($73.00) | €2.52 (PAYG) |
| Compute (2 vCPU, 4 GiB) | €150 to €300 (assumed, redundant nodes) | ~€50 ($54.75 incl. ALB) | €36.50 |
| Storage 20 GiB | included | ~€2 ($1.90) | €1.60 |
| Egress 50 GiB | included | €0 (free allowance) | €0.50 |
| Monthly total (net) | €150 to €300 + labor | ~€118 | €41.12 (€48.93 gross) |
| Ops overhead | High | Medium | None for the cluster |
| Provider and data location | Your choice | US corporation, EU region available | German provider, servers in Germany, backups in Finland (EU) |
All conversions use 1 USD = 0.91 EUR (ECB reference rate of 20 November 2026). All amounts net, with the gross amount shown next to ITSH. The self-hosted column matches the assumed infrastructure figure above: a redundant control plane needs several nodes, a single VPS won’t do.
Prices as of 22 November 2026. AWS prices per the public price list (EKS, EC2, VPC), ITSH prices per the pricing calculator.
Calculate the cost of your workload →
What’s Included in an ITSH Namespace?
The price includes services you would have to build yourself on your own cluster:
- GitOps with ArgoCD: In the portal, you switch on deploying from a Git repository for your namespace, and from then on a deployment is a push. The repository must be readable without credentials, so in practice public. See Deploying from Git.
- Daily backups: Volumes and namespace config are backed up daily. Restores currently run through support rather than self-service. This does not replace your own backups, and databases need a backup of their own (how to set one up).
- Autoscaling: HorizontalPodAutoscaler, KEDA, and scale-to-zero work in the namespace, and you set them up yourself. Each hour is billed at the higher of requested and actually used resources.
- Visibility into your workloads: The customer portal shows pods, logs, events, and your resource usage, and the platform itself is monitored around the clock. Direct Prometheus access, your own ServiceMonitors or PrometheusRules, and Grafana access are not part of it.
- TLS certificates: Automatic once your HTTPRoute carries the
cert-manager.io/cluster-issuerannotation.
Keeping Costs Under Control
Requests are the least you pay for CPU and memory. Three tools keep the bill small:
- Sizing suggestions: The Workloads view in the portal compares what a workload requests with its usage over the last seven days and suggests a request that fits, with the monthly saving next to it.
- Spending cap: You set a spending cap for the namespace in the portal. Do it before experimenting with anything that scales.
- Scale-to-zero: An HTTP service can drop to zero replicas when nothing calls it and wake on the next request. While it sleeps, it costs nothing in CPU and memory. Storage and the base fee continue.
How billing works in detail is covered in Billing and quotas.
When Does Each Model Make Sense?
Self-hosted is worth it if you already have a DevOps team and need full control over the cluster, for example due to specific compliance requirements or GPU workloads.
AWS/GKE/AKS makes sense if you need a hyperscaler’s other services, run in several regions, or want your own cluster with admin rights.
As an alternative to AWS EKS, a managed namespace with ITSH fits if you:
- Want to run containers in production without managing an entire cluster
- Value a German provider with servers in Germany
- Have a limited budget (startups, agencies, SMBs)
- Don’t employ a dedicated DevOps team (Container Deployment Without a DevOps Team)
Classic web hosting may be enough for you, too. How to tell is covered in Kubernetes vs. Shared Hosting.
Worked Example: SaaS Startup with Three Microservices
A typical setup of an API backend, a worker service, and a frontend container:
# Estimated resource requirements
API: 0.5 vCPU, 512 MiB RAM
Worker: 0.5 vCPU, 1 GiB RAM
Frontend: 0.25 vCPU, 256 MiB RAM
Storage: 10 GiB (database PVC)
Egress: 30 GiB/month
Monthly cost on ITSH (PAYG, net):
- Base fee: €2.52
- 1.25 vCPU × 730 h × €0.015: €13.69
- 1.75 GiB RAM × 730 h × €0.005: €6.39
- 10 GiB Storage: €0.80
- 30 GiB Egress: €0.30
- Total: €23.70/month net (€28.20 gross)
The same setup on AWS EKS: at least $110/month net (about €100), more than four times the price.
Conclusion
Kubernetes doesn’t have to be expensive. Affordable Kubernetes hosting exists if you don’t need your own cluster: for containers that should run reliably without the full feature catalog of a hyperscaler, a managed namespace is significantly cheaper.
From €3/month base fee plus usage (gross), you can get going on ITSH without a cluster setup, with a German provider and servers in Germany.
Calculate the cost of your workload →
Or try it for free: Free tier namespace →